Data Used in this report

Large-scale filtering

Data used to create these indices were taken from the GSMFC website in March of 2019. Data was then limited to 40 ft trawls, and any tows with no operation codes (indicating a bad station) or an op code of “W” (water tow, aka no catch but otherwise fine). Data was also limited to summer and fall groundfish cruises specifically. Stations with no Stat zone or tow speed records were also removed.

A survey design change ocurred in 2008 between the summer and fall groundfish surveys. This change expanded the area sampling into Florida waters, an area previously unsampled, and changed from a dapth stratified sampling design where multiple tows longer than 30min were common. The new (current) sampling design covers the entire Gulf from Brownsville TX to the Florida Keyes, and each station consists of a single 30 minute tow.

It was recommended to me by the analyst group at the Pascagoula Laboratory to use groundfish data from 2010* forward for consistency as the design change was not uniformly applied until then, with some state partners sampling differently.

Additional filters from data exploration

Based on exploration of the data, and in the interest of looking at Gulfwide or atleast East/West trends several additional filters were added. Data was limited to years after 2009 for consistency in the sampling methods and for a Gulfwide coverage. Additionally statzones 0 (not a real zone), 1 (The Keyes), 12 (MS Sound, inside Chandy), and 22 (Brownsville/MX) were not included due to a lack of observations. Stat zones 6 & 9 were also excluded from the model, as they have no observed C. sapidus catch for the timeseries and thus have no expected catch.

There was an initial concern with abnormally high catch values being potential mis-identifications. This is less of a concern with the reduced time-series

Data Set Summary

  • Total Number of Stations = 6254

  • Starting in 2009 and continuing through 2018

  • Total Sapidus Caught was 2280

  • Overall Frequency of Occurrence was 13.4793732

Factor Summaries

Yearly Summaries

Overall Mean Catch: 0.3645667 Stations with no catch: 5411 / 6254 or 86.5206268

Survey_Year Stations Missing Obs % Occurrence Mean Catch SD Overdispersed
2009 983 0 0.1576806 0.3601221 1.2885667 yes
2010 700 0 0.1571429 0.4271429 1.5527373 yes
2011 549 0 0.2021858 0.6484517 2.2777691 yes
2012 589 0 0.0916808 0.1782683 0.7050376 yes
2013 506 0 0.0790514 0.1620553 0.8309740 yes
2014 685 0 0.1138686 0.3795620 3.7065837 yes
2015 700 0 0.1314286 0.3728571 1.7779786 yes
2016 548 0 0.1423358 0.3521898 1.3541374 yes
2017 603 0 0.1293532 0.4560531 1.8992853 yes
2018 391 0 0.1202046 0.2429668 1.0050107 yes

Yearly Catch Plot


Yearly Frequency of Occurrence


Yearly Side-by


Stat Zone Summaries

StatZone Stations Missing Obs % Occurrence Mean Catch SD Overdispersed
1 5 0 0.0000000 0.0000000 0.0000000 no
2 125 0 0.0160000 0.0160000 0.1259800 no
3 399 0 0.0100251 0.0100251 0.0997472 no
4 443 0 0.0067720 0.0135440 0.1774552 yes
5 396 0 0.0075758 0.0101010 0.1228315 yes
6 395 0 0.0000000 0.0000000 0.0000000 no
7 245 0 0.0122449 0.0163265 0.1559568 yes
8 268 0 0.0559701 0.2126866 1.4047497 yes
9 191 0 0.0000000 0.0000000 0.0000000 no
10 173 0 0.0462428 0.0462428 0.2106200 no
11 366 0 0.2349727 0.5901639 1.7253347 yes
12 17 0 0.1764706 0.4117647 1.0641207 yes
13 129 0 0.3875969 1.3333333 3.6407703 yes
14 285 0 0.2491228 0.5578947 1.4465777 yes
15 339 0 0.2684366 0.5988201 1.6636023 yes
16 440 0 0.2431818 0.7022727 2.1871429 yes
17 549 0 0.2386157 0.6885246 2.0216609 yes
18 506 0 0.2055336 0.5474308 1.9280944 yes
19 353 0 0.2322946 0.8441926 5.1902527 yes
20 361 0 0.1274238 0.2908587 1.2610179 yes
21 268 0 0.1268657 0.2649254 0.9163318 yes

Stat Zone Catch Plot


Stat Zone Frequency of Occurrence


Stat Zone Side-by


Region Summary

region Stations Missing Obs % Occurrence Mean Catch SD Overdispersed
Texas 1488 0 0.1787634 0.5047043 2.8678548 yes
Louisiana 1743 0 0.2581756 0.7005164 2.0925308 yes
MS Bight 556 0 0.1744604 0.4154676 1.4376075 yes
Florida 2467 0 0.0121605 0.0312120 0.4801817 yes

Regional Catch Plot


Regional Frequency of Occurrence


Regional Side-by


Region - Season

region Season Stations Missing Obs % Occurrence Mean Catch SD Overdispersed
Texas Summer 827 0 0.2394196 0.7255139 3.6214139 yes
Texas Fall 661 0 0.1028744 0.2284418 1.4069661 yes
Louisiana Summer 977 0 0.3387922 0.9713408 2.3900403 yes
Louisiana Fall 766 0 0.1553525 0.3550914 1.5715673 yes
MS Bight Summer 315 0 0.2444444 0.6571429 1.8502637 yes
MS Bight Fall 241 0 0.0829876 0.0995851 0.3512475 yes
Florida Summer 1540 0 0.0162338 0.0467532 0.6046354 yes
Florida Fall 927 0 0.0053937 0.0053937 0.0732833 no

Region - Season Catch


Region - Season FOO


Region - Season Side-by


Seasonal Summary

Survey_Year Season Stations Missing Obs % Occurrence Mean Catch SD Overdispersed
2009 Summer 539 0 0.2393321 0.5918367 1.6765406 yes
2009 Fall 444 0 0.0585586 0.0788288 0.3499145 yes
2010 Summer 386 0 0.2020725 0.6217617 1.9611666 yes
2010 Fall 314 0 0.1019108 0.1878981 0.7411114 yes
2011 Summer 329 0 0.2401216 0.7294833 1.9637911 yes
2011 Fall 220 0 0.1454545 0.5272727 2.6799069 yes
2012 Summer 391 0 0.1023018 0.2020460 0.7179714 yes
2012 Fall 198 0 0.0707071 0.1313131 0.6781052 yes
2013 Summer 313 0 0.1118211 0.2300319 0.9894276 yes
2013 Fall 193 0 0.0259067 0.0518135 0.4534696 yes
2014 Summer 364 0 0.1236264 0.5274725 5.0164263 yes
2014 Fall 321 0 0.1028037 0.2118380 0.8760619 yes
2015 Summer 377 0 0.1803714 0.5172414 1.8794365 yes
2015 Fall 323 0 0.0743034 0.2043344 1.6385438 yes
2016 Summer 354 0 0.1638418 0.4435028 1.5584641 yes
2016 Fall 194 0 0.1030928 0.1855670 0.8435049 yes
2017 Summer 322 0 0.1987578 0.7950311 2.5286348 yes
2017 Fall 281 0 0.0498221 0.0676157 0.3766170 yes
2018 Summer 284 0 0.1232394 0.2746479 1.1318977 yes
2018 Fall 107 0 0.1121495 0.1588785 0.5346403 yes

Seasonal Catch Plot


Seasonal Frequency of Occurrence


Seasonal Side-by


Combining Figures

1x2

TableGrob (2 x 1) "arrange": 2 grobs
  z     cells    name           grob
1 1 (1-1,1-1) arrange gtable[layout]
2 2 (2-2,1-1) arrange gtable[layout]

TableGrob (2 x 1) "arrange": 2 grobs
  z     cells    name           grob
1 1 (1-1,1-1) arrange gtable[layout]
2 2 (2-2,1-1) arrange gtable[layout]

TableGrob (2 x 1) "arrange": 2 grobs
  z     cells    name           grob
1 1 (1-1,1-1) arrange gtable[layout]
2 2 (2-2,1-1) arrange gtable[layout]

3x2 A.

TableGrob (3 x 2) "arrange": 6 grobs
  z     cells    name           grob
1 1 (1-1,1-1) arrange gtable[layout]
2 2 (1-1,2-2) arrange gtable[layout]
3 3 (2-2,1-1) arrange gtable[layout]
4 4 (2-2,2-2) arrange gtable[layout]
5 5 (3-3,1-1) arrange gtable[layout]
6 6 (3-3,2-2) arrange gtable[layout]

3x2 B.

TableGrob (2 x 3) "arrange": 6 grobs
  z     cells    name           grob
1 1 (1-1,1-1) arrange gtable[layout]
2 2 (1-1,2-2) arrange gtable[layout]
3 3 (1-1,3-3) arrange gtable[layout]
4 4 (2-2,1-1) arrange gtable[layout]
5 5 (2-2,2-2) arrange gtable[layout]
6 6 (2-2,3-3) arrange gtable[layout]

Index of Abundance

Negative Binomial IOA

A negative binomial generalized linear model was used to estimate an index of abundance for the time series. The factors considered in the model were Year, Season, and Stat Zone.

Model Selection

Method : Move from the simplest model of catch and year to the most complex with year, season, statzone, depth, temp, and salinity.

Most parsimonious model (AIC) :

year + season + stat zone

OR

year + season + region

significant predictors:

term estimate std.error statistic p.value
(Intercept) -4.1399866 0.7335176 -5.6440182 0.0000000
Year_f2010 0.4622160 0.1654732 2.7932992 0.0052173
Year_f2011 0.8056565 0.1664822 4.8392942 0.0000013
Year_f2012 -0.4183861 0.1939715 -2.1569461 0.0310099
Year_f2013 -0.4808636 0.2086077 -2.3051100 0.0211604
Year_f2014 0.6173601 0.1710930 3.6083301 0.0003082
Year_f2015 0.6073110 0.1699604 3.5732501 0.0003526
Year_f2016 0.2421701 0.1801011 1.3446341 0.1787434
Year_f2017 0.3667771 0.1771867 2.0700037 0.0384520
Year_f2018 0.1321023 0.2257196 0.5852497 0.5583799
SeasonFall -1.2595060 0.0937274 -13.4379691 0.0000000
StatZone3 -0.5327310 0.8882528 -0.5997515 0.5486719
StatZone4 -0.1764824 0.8419341 -0.2096154 0.8339678
StatZone5 -0.4477110 0.8907543 -0.5026201 0.6152314
StatZone7 0.1134053 0.8917659 0.1271694 0.8988063
StatZone8 2.5723013 0.7534550 3.4140080 0.0006401
StatZone10 1.2341286 0.8237717 1.4981441 0.1340958
StatZone11 3.7437210 0.7388302 5.0670925 0.0000004
StatZone13 4.7094645 0.7570346 6.2209371 0.0000000
StatZone14 3.9054392 0.7417412 5.2652315 0.0000001
StatZone15 3.8056239 0.7392212 5.1481532 0.0000003
StatZone16 3.8604000 0.7354000 5.2493880 0.0000002
StatZone17 3.6891521 0.7337145 5.0280485 0.0000005
StatZone18 3.6498597 0.7348862 4.9665649 0.0000007
StatZone19 3.8149649 0.7385118 5.1657465 0.0000002
StatZone20 3.1150148 0.7417669 4.1994527 0.0000268
StatZone21 2.9070554 0.7492185 3.8801169 0.0001044

qqplot

I’ve seen worse, obviously not perfect though.

Predicting abundance indices (Statzone - mod)

Generate the data for a timeline using estimated marginal means. Then plot that timeline with Confidence interval.

Observed vs. Predicted

Model prediction (black) & observed mean values (blue), both standardized by overall mean

Compare to state catch

State landings data taken from GEDAR assesment and covers years 2000-2011. The values were divided by the overall mean for the timeseries for each state then plotted. They aren’t weighted according to which State catches more so this plot isn’t great.

Pretty noisy, and gives you a sense that either the landings reflect the population poorly, or that the population is resilient to some.

Looking at the other factors

Statzones start counting at 1 (FL Keyes) and continue through brownsville Texas (22), 12 is around the Chandeleurs, and 13 is Terrebonne Bay and the MS Birdfoot.

NB GLM - Year + Season + Region FINAL

Update 10/3/2019

Need To account for catch rates not being discrete with either an offset or with weights

Formula for using an offset:
Log(Catch)y, s, r  ∼ λ + λyYear + λsSeasonλrRegion + Log(Towtime)

Link to Reference: https://stats.stackexchange.com/questions/66791/where-does-the-offset-go-in-poisson-negative-binomial-regression/66878#66878

Link to formula for GLM https://newonlinecourses.science.psu.edu/stat504/node/216/

Use emmeans on year + season + region

term estimate std.error statistic p.value
(Intercept) -3.9233196 0.1294643 -30.3042577 0.0000000
Year_f2010 0.2493999 0.1683396 1.4815290 0.1384657
Year_f2011 0.6976250 0.1693122 4.1203466 0.0000378
Year_f2012 -0.5560169 0.1971552 -2.8201997 0.0047994
Year_f2013 -0.5026238 0.2078878 -2.4177655 0.0156161
Year_f2014 0.4309158 0.1736025 2.4821987 0.0130574
Year_f2015 0.6461253 0.1691059 3.8208326 0.0001330
Year_f2016 0.0605462 0.1843363 0.3284552 0.7425675
Year_f2017 0.2844148 0.1790591 1.5883852 0.1121993
Year_f2018 0.0478424 0.2244907 0.2131152 0.8312371
SeasonFall -1.2988110 0.0956061 -13.5850246 0.0000000
regionLouisiana 0.4071318 0.1023161 3.9791566 0.0000692
regionMS Bight -0.0857449 0.1501854 -0.5709272 0.5680490
regionFlorida -2.8634887 0.1517673 -18.8676315 0.0000000

Table of emmeans

Year_f Season region response SE df asymp.LCL asymp.UCL
2009 Summer Texas 0.5915148 0.0765800 Inf 0.4589500 0.7623700
2010 Summer Texas 0.7590644 0.1115539 Inf 0.5690941 1.0124490
2011 Summer Texas 1.1883387 0.1766205 Inf 0.8880304 1.5902034
2012 Summer Texas 0.3392271 0.0602380 Inf 0.2395185 0.4804431
2013 Summer Texas 0.3578317 0.0677764 Inf 0.2468619 0.5186849
2014 Summer Texas 0.9101437 0.1408446 Inf 0.6720268 1.2326316
2015 Summer Texas 1.1286889 0.1684645 Inf 0.8424161 1.5122440
2016 Summer Texas 0.6284352 0.1035753 Inf 0.4549567 0.8680623
2017 Summer Texas 0.7861137 0.1252708 Inf 0.5752313 1.0743065
2018 Summer Texas 0.6205021 0.1286445 Inf 0.4133038 0.9315735
2009 Fall Texas 0.1613984 0.0225483 Inf 0.1227386 0.2122351
2010 Fall Texas 0.2071153 0.0320308 Inf 0.1529580 0.2804479
2011 Fall Texas 0.3242454 0.0499073 Inf 0.2398050 0.4384190
2012 Fall Texas 0.0925602 0.0174351 Inf 0.0639863 0.1338939
2013 Fall Texas 0.0976365 0.0195894 Inf 0.0658917 0.1446752
2014 Fall Texas 0.2483382 0.0398646 Inf 0.1813026 0.3401599
2015 Fall Texas 0.3079696 0.0479826 Inf 0.2269279 0.4179532
2016 Fall Texas 0.1714723 0.0301160 Inf 0.1215336 0.2419312
2017 Fall Texas 0.2144959 0.0356782 Inf 0.1548225 0.2971692
2018 Fall Texas 0.1693078 0.0376230 Inf 0.1095281 0.2617146
2009 Summer Louisiana 0.8887522 0.1112690 Inf 0.6953652 1.1359218
2010 Summer Louisiana 1.1404959 0.1634662 Inf 0.8611757 1.5104128
2011 Summer Louisiana 1.7854816 0.2596853 Inf 1.3426237 2.3744140
2012 Summer Louisiana 0.5096894 0.0886677 Inf 0.3624324 0.7167772
2013 Summer Louisiana 0.5376429 0.0998734 Inf 0.3735712 0.7737745
2014 Summer Louisiana 1.3674928 0.2051141 Inf 1.0191801 1.8348441
2015 Summer Louisiana 1.6958575 0.2465628 Inf 1.2753581 2.2550001
2016 Summer Louisiana 0.9442252 0.1499758 Inf 0.6916321 1.2890685
2017 Summer Louisiana 1.1811375 0.1848247 Inf 0.8691690 1.6050800
2018 Summer Louisiana 0.9323058 0.1892990 Inf 0.6262202 1.3880007
2009 Fall Louisiana 0.2425014 0.0325774 Inf 0.1863651 0.3155469
2010 Fall Louisiana 0.3111912 0.0466564 Inf 0.2319573 0.4174904
2011 Fall Louisiana 0.4871794 0.0728657 Inf 0.3633941 0.6531306
2012 Fall Louisiana 0.1390718 0.0255788 Inf 0.0969800 0.1994327
2013 Fall Louisiana 0.1466991 0.0287854 Inf 0.0998629 0.2155018
2014 Fall Louisiana 0.3731287 0.0577261 Inf 0.2755311 0.5052969
2015 Fall Louisiana 0.4627250 0.0697957 Inf 0.3442951 0.6218920
2016 Fall Louisiana 0.2576375 0.0435148 Inf 0.1850298 0.3587374
2017 Fall Louisiana 0.3222805 0.0523402 Inf 0.2344197 0.4430717
2018 Fall Louisiana 0.2543853 0.0552901 Inf 0.1661442 0.3894922
2009 Summer MS Bight 0.5429090 0.0862470 Inf 0.3976531 0.7412243
2010 Summer MS Bight 0.6966908 0.1262825 Inf 0.4883722 0.9938691
2011 Summer MS Bight 1.0906909 0.2001007 Inf 0.7612693 1.5626622
2012 Summer MS Bight 0.3113522 0.0647840 Inf 0.2070808 0.4681274
2013 Summer MS Bight 0.3284281 0.0712170 Inf 0.2147158 0.5023617
2014 Summer MS Bight 0.8353556 0.1559391 Inf 0.5793944 1.2043937
2015 Summer MS Bight 1.0359426 0.1880720 Inf 0.7257764 1.4786608
2016 Summer MS Bight 0.5767956 0.1130683 Inf 0.3927916 0.8469965
2017 Summer MS Bight 0.7215174 0.1373168 Inf 0.4968771 1.0477185
2018 Summer MS Bight 0.5695144 0.1308793 Inf 0.3629872 0.8935484
2009 Fall MS Bight 0.1481360 0.0248930 Inf 0.1065671 0.2059198
2010 Fall MS Bight 0.1900963 0.0357761 Inf 0.1314552 0.2748967
2011 Fall MS Bight 0.2976016 0.0560752 Inf 0.2057069 0.4305482
2012 Fall MS Bight 0.0849543 0.0185130 Inf 0.0554235 0.1302197
2013 Fall MS Bight 0.0896136 0.0203650 Inf 0.0574031 0.1398983
2014 Fall MS Bight 0.2279318 0.0437917 Inf 0.1564108 0.3321568
2015 Fall MS Bight 0.2826632 0.0530329 Inf 0.1956897 0.4082916
2016 Fall MS Bight 0.1573822 0.0323895 Inf 0.1051422 0.2355776
2017 Fall MS Bight 0.1968704 0.0387482 Inf 0.1338585 0.2895441
2018 Fall MS Bight 0.1553954 0.0378903 Inf 0.0963582 0.2506040
2009 Summer Florida 0.0337573 0.0057339 Inf 0.0241984 0.0470923
2010 Summer Florida 0.0433193 0.0077698 Inf 0.0304796 0.0615678
2011 Summer Florida 0.0678177 0.0121820 Inf 0.0476917 0.0964368
2012 Summer Florida 0.0193595 0.0039983 Inf 0.0129150 0.0290195
2013 Summer Florida 0.0204212 0.0044363 Inf 0.0133403 0.0312607
2014 Summer Florida 0.0519413 0.0092936 Inf 0.0365773 0.0737588
2015 Summer Florida 0.0644135 0.0111488 Inf 0.0458828 0.0904282
2016 Summer Florida 0.0358644 0.0069272 Inf 0.0245614 0.0523688
2017 Summer Florida 0.0448630 0.0084294 Inf 0.0310423 0.0648370
2018 Summer Florida 0.0354116 0.0079605 Inf 0.0227927 0.0550169
2009 Fall Florida 0.0092109 0.0017203 Inf 0.0063874 0.0132825
2010 Fall Florida 0.0118199 0.0022970 Inf 0.0080761 0.0172993
2011 Fall Florida 0.0185045 0.0035662 Inf 0.0126833 0.0269973
2012 Fall Florida 0.0052823 0.0011797 Inf 0.0034098 0.0081833
2013 Fall Florida 0.0055721 0.0013050 Inf 0.0035210 0.0088179
2014 Fall Florida 0.0141725 0.0027302 Inf 0.0097157 0.0206738
2015 Fall Florida 0.0175756 0.0032988 Inf 0.0121660 0.0253907
2016 Fall Florida 0.0097858 0.0020588 Inf 0.0064791 0.0147802
2017 Fall Florida 0.0122411 0.0024742 Inf 0.0082372 0.0181913
2018 Fall Florida 0.0096623 0.0023708 Inf 0.0059734 0.0156293

Full Term Plot

Interaction plot Year & Season

Interaction Plot Year & Region

Interaction Plot Season & Region

Effort Mapping

These maps are made using only the model data so years 2010-2018 and only the stat zones with enough samples and some catch. The gaps around Pensacola and Tampa are those areas I took out because there was no catch and they made the models confidence intervals absurd.

Stations w/ Stat Zone Numbers

Effort Heatmap

Catch Heatmap

Comments

I have polygons for the stat zones and was thinking I would make a map with those polygons shaded by predicted catch or with the mean catch visible within them. That file is too big for my laptop to process though so I couldn’t add it here. There’s really three big things with the data.

  1. years are fairly variable with little consistency or things I would consider long-term trends.
  2. there is pretty consistently more crabs in the Western half of the Gulf, Particularly off Louisiana. 3. There are less crabs in the Fall Survey than the Summer.

If the goal was to explain why this is so I’d say that its likely the difference in bottom types in these two offshore areas sand/hard bottom vs. mud and then coupled with more crabs inshore off Louisiana so you are just starting with more crabs to begin with. But, for this paper I’m thinking we point out

  • Offshore population, as predicted from SEAMAP is noisy, and that it would make sense to split it East and West based on the locations where we see them and how different the habitats become
  • Point out that we see less as we sampling moves East and the largest offshore population in off Louisiana
  • Point out that there are many fewer in the Fall